A ferry-ticket office has a still of a pass, a spoken request, and a typed note. Something must decide which encoder each stream enters and when those streams may meet. What is that routing of data types called?

Something that decides which encoder a ferry-ticket still, spoken request, and typed note each enter—and when those streams may meet—is modality orchestration: scheduling and routing modalities through encoders and fusion. That routing is not identical to late fusion’s averaging of final labels. Agent orchestration is about sequencing acts/tools/roles, not defined as averaging pixel tensors. A context-window hyperparameter limits text capacity; it is not the routing role. Exam caveat: modality orchestration = route by data type through encoders/fusion. Operational check: draw a router that sends pass images, speech, and notes to the correct encoders and gates when fusion occurs; label that diagram modality orchestration. Implement a ferry-office router test that sends each modality to the correct encoder stub. Modality orchestration routes ferry stills, speech, and notes into encoders and fusion schedules. Implement ferry-office router tests that assert each of still, speech, and note enters the matching encoder stub and that fusion is scheduled explicitly. Close q128 with a municipal evidence note that records why the correct mechanism fits this civic multimodal scenario, why each of the three distractors fails by a distinct concept, and which operational check was observed in logs or dataset cards before promotion. File that note with the run packet so later maintenance cannot silently revive a rejected framing.